DoorDash Data Analyst Interview: SQL, Python, Business Insights, and A/B Testing Full Assessment

Background

I have 2 years of data analysis experience, previously working at an e-commerce company doing user growth analytics. DoorDash's Data Analyst position was one I really wanted—DoorDash's data-driven culture is well-known across the tech industry. They say even the color of a button is decided through A/B testing. I applied for the Data Analyst position at DoorDash, and the entire interview process took about two and a half weeks. The pace was very fast: three technical rounds plus an HR round, with almost no waiting time in between.

During preparation, I focused on reviewing SQL window functions (an area I wasn't very strong in), advanced Python pandas usage, statistical principles of A/B testing, and business case analysis frameworks. Honestly, DoorDash's interviews are very practical—you can't pass by just memorizing textbook answers. Every question is closely tied to real business scenarios.

Interview Process Review

Round 1: SQL Window Functions + Python pandas

My first interviewer was a woman, likely a senior analyst on the data team. After a brief self-introduction, we jumped straight into writing SQL.

The first SQL question caught me off guard: Given a user orders table, find the average amount of each user's 3 most recent orders, using window functions. I wrote a solution using ROW_NUMBER() OVER (PARTITION BY user_id ORDER BY order_date DESC), then filtered rn <= 3 with a subquery before aggregating. The interviewer said the approach was fine but followed up with: What if a user has fewer than 3 orders? What's the difference between ROWS BETWEEN and RANGE BETWEEN? I explained that fewer than 3 orders would be handled naturally, then described how ROWS offsets by physical rows while RANGE offsets by logical values.

The second SQL question was harder: Calculate the next-day retention rate for daily new users in a single SQL query. I had practiced similar problems before, so I used a self-join + conditional aggregation approach. First, find each user's first login date as their new user date, then LEFT JOIN the next day's login records, and finally calculate the retention rate. The interviewer followed up: If the data volume is very large, how would you optimize this query? I mentioned partitioned tables, index optimization, and using CTEs instead of subqueries for readability.

Python section: How do you implement multi-level aggregation using pandas' groupby and agg chaining? I wrote code on the spot demonstrating groupby + agg usage, including custom aggregation functions. The interviewer then asked a practical question: If you have 100 million rows and pandas can't handle it, what do you do? I mentioned chunk-based reading, Dask for parallel computing, and Spark as alternatives. The interviewer seemed satisfied and followed up with: Compare the applicable scenarios for Dask vs Spark.

Round 1 ended with an open-ended question: If your boss asks you to analyze why last week's GMV dropped by 20%, how would you approach it? I outlined a breakdown approach from four dimensions: time, channel, category, and user, then mentioned the need to exclude external factors like holidays. The interviewer said "clear thinking" and gave me a Round 2 opportunity.

Round 1 lasted about 55 minutes, with SQL taking up the majority of time and Python being relatively less.

Round 2: A/B Testing + Business Cases

The second interviewer was the data team lead, and the style was more business-discussion oriented. The opening question made me nervous: What does the p-value in A/B testing mean? If the p-value is less than 0.05, does that mean the experiment is definitely effective? I explained that the p-value is the probability of observing the current or more extreme results under the null hypothesis, and that p < 0.05 only means statistical significance, not necessarily business significance. The interviewer followed up: What is statistical power? Why is it important? I explained that power is the probability of not making a Type II error, and insufficient power leads to false negatives.

A/B test design question: If you were testing a new recommendation algorithm, how would you design the experiment? What factors need to be considered? I detailed the approach from several aspects: experimental grouping (randomization, stratification), sample size calculation (based on MDE and power), experiment duration (considering periodic effects), and metric selection (primary metrics + guardrail metrics). The interviewer followed up: If the treatment and control groups have unequal sample sizes, what's the impact? I explained the asymmetry of variance and why equal splitting is generally recommended.

Business case: How would you evaluate the effectiveness of a major promotional campaign? This was a big question. I outlined an analysis framework from three dimensions: incremental effects (using DID or PSM methods), long-term impact (user LTV changes), and spillover effects (impact on non-promoted items). The interviewer was interested in the PSM part and followed up: What's the principle behind PSM? How is the propensity score calculated? What matching methods are available? I detailed logistic regression for propensity score calculation, nearest-neighbor matching and kernel matching methods, and balance checking steps.

Round 2 also included a very practical question: If the A/B test results contradict business intuition, what do you do? I said I'd first check if the experimental design had issues (successful randomization, no spillover effects), then check data quality, and finally, if everything checks out, respect the data while understanding the underlying reasons. The interviewer seemed to appreciate this answer.

Round 3: Data-Driven Decision Making + HR Round

Round 3 was a cross-team interview with an interviewer from another data team. The style was more project-focused. They asked about the most impactful analysis project I'd worked on. I chose a user churn prediction project and detailed the entire process from feature engineering to model selection to business implementation. The interviewer followed up: How do you measure the business value of this analysis project? I explained quantifying through retention rate improvement after intervention, and how to attribute using A/B testing.

Data-driven decision question: If a product manager proposes a feature, but your data analysis suggests it won't have a positive effect, how do you communicate this? I shared a real experience where I used data to convince a PM to drop a feature. The key was translating data analysis results into business language rather than just throwing numbers at them. The interviewer nodded.

The HR round was fairly standard—career planning, why DoorDash, salary expectations, etc. The HR was candid about the fast pace but emphasized the growth opportunities.

Interview Questions Summary

Round 1:

1. SQL window functions: Average amount of each user's 3 most recent orders

2. Difference between ROWS BETWEEN and RANGE BETWEEN

3. SQL: Next-day retention rate for daily new users

4. Large data volume SQL query optimization

5. Pandas multi-level aggregation with chaining

6. Pandas alternatives for large data (Dask vs Spark)

7. Open-ended: Analyzing a 20% GMV drop

Round 2:

1. P-value meaning and limitations in A/B testing

2. Statistical power concept and importance

3. Recommendation algorithm A/B test design

4. Impact of unequal treatment/control group sizes

5. Promotional campaign effectiveness evaluation

6. PSM principles, propensity score calculation, and matching methods

7. Handling contradictions between A/B results and business intuition

Round 3:

1. Most impactful analysis project deep dive (user churn prediction)

2. How to measure business value of analysis projects

3. Data analysis communication strategies with product managers

4. Career planning, why DoorDash

Key Takeaways

1. SQL window functions are mandatory: DoorDash's data analyst interviews will definitely test window functions, and not just simple ROW_NUMBER. Expect ROWS/RANGE, sliding windows, cumulative aggregation, and other advanced usage. Practice all window function patterns thoroughly.

2. Understand A/B testing statistical principles: Don't just know how to calculate p-values—understand the complete hypothesis testing framework, statistical power, multiple comparison corrections, etc. DoorDash values depth of understanding in experimental methodology.

3. Have a framework for business cases: Don't just talk stream-of-consciousness. Use a structured analysis framework. My go-to framework: Define problem → Break down dimensions → Form hypotheses → Test hypotheses → Draw conclusions. Interviewers value your analytical logic.

4. Python shouldn't stop at pandas: While pandas suffices for daily work, interviews will ask about alternatives for big data scenarios. At minimum, understand the basic concepts of Dask and Spark.

5. Communication skills matter as much as technical skills: Round 3 and the HR round both assess how you communicate data insights to non-technical people. Data analysis isn't just writing SQL—it's using data to influence decisions.

FAQ

Q: How deep is the SQL assessment in DoorDash data analyst interviews?

A: Very deep. Not just simple queries—expect window functions and complex aggregations. I recommend practicing SQL problems on LeetCode, especially Hard-level ones.

Q: Can I interview for a data analyst role without A/B testing experience?

A: It's difficult. DoorDash data analysts work with A/B testing almost every day—it's a core competency. Without practical experience, at minimum, you need to understand the statistical principles thoroughly.

Q: How should I prepare for business cases?

A: I recommend studying real internet business analysis cases and understanding common frameworks (AARRR, RFM, attribution analysis, etc.). During interviews, don't rush to give answers—first confirm the problem boundaries.

Q: What's the interview pace like?

A: Very fast—two and a half weeks for the complete process. DoorDash is highly efficient, with results within 1-2 days after each round. The fast pace means limited preparation time, so prepare thoroughly in advance.

Q: What's the work intensity like for DoorDash data analysts?

A: HR was candid about the fast pace. From what I understand, DoorDash does have significant overtime, but the data analyst role is relatively better than engineering roles. The data-driven culture is genuine—every decision is backed by data.